A

(abmtraviss©t

K

1

isoAotn

t2

M

1C

Available online at www.sciencedirect.com

Journal of Applied Researchand Technology

www.jart.ccadet.unam.mxJournal of Applied Research and Technology 15 (2017) 1–13

Original

Simplified performance estimation of ISM-band, OFDM-based WSNsaccording to the sensitivity/SINAD parameters

Pierre van Rhyn ∗, Gerhard P. HanckeDepartment of Electrical, Electronic and Computer Engineering, University of Pretoria, Private Bag X20, Hatfield 0028, Lynnwood Road, Pretoria, South Africa

Received 13 January 2016; accepted 4 October 2016Available online 13 January 2017

bstract

A novel method is proposed to estimate committed information rate (CIR) variations in typical orthogonal frequency division multiplexingOFDM) wireless local area networks (WLANs) that are applied in wireless sensor networks (WSNs) and operate within the industrial, scientificnd medical (ISM) frequency bands. The method is based on the observation of a phenomenon of which the significance has not previouslyeen recognized nor documented; here termed the service level differential zone (SLDZ). This method, which conforms to the ITU-T Y.1564 testethodology, provides the means to set a committed information rate (CIR) reference for IEEE 802.11a/g/n OFDM systems in terms of committed

hroughput bandwidth between a test node and an access point (AP) at a specific range. An analytical approach is presented to determine theelationship between the maximum operating range (in metres) of a wireless sensor network for a specific committed throughput bandwidth,nd its link budget (in dB). The most significant contributions of this paper are the analytical tools to determine wireless network capabilities,ariations and performance in a simplified method, which does not require specialized measurement equipment. With these it becomes possible forndustrial technicians and engineers (who are not necessarily information technology (IT) network experts) to field analyze OFDM WLANs and

o qualify their performance in terms of Y.1564 specified service level agreement (SLA) requirements, as well as in terms of the widely acceptableensitivity/SINAD parameters.

2016 Universidad Nacional Autónoma de México, Centro de Ciencias Aplicadas y Desarrollo Tecnológico. This is an open access article underhe CC BY-NC-ND license (http://creativecommons.org/licenses/by-nc-nd/4.0/).

form

Ti(2

tpFa‘iac

eywords: Mobile; Wireless; Sensor networks; Data throughput; Committed in

. Introduction

WSNs consist of sensor nodes that are placed in close prox-mity to, or within the phenomena they observe, according topecific measurement requirements, whilst communicating withther nodes or network access points (Hou & Bergmann, 2010).

true-to-type WSN sensor node consists of a suitable sensorr transducer, a microprocessor to execute trivial logic opera-ions or complex programmes, an RF transceiver to access theetwork and a suitable power supply (Gungor & Hancke, 2009).

WSNs have been successfully applied in environmental andemperature monitoring (Carullo, Corbellini, Parvis, & Vallan,009), machine monitoring (Flammini, Marioli, Sisinni, &

∗ Corresponding author.E-mail address: vanrhyn@mweb.co.za (P. van Rhyn).

Peer Review under the responsibility of Universidad Nacional Autónoma deéxico.

cct

f

http://dx.doi.org/10.1016/j.jart.2016.10.002665-6423/© 2016 Universidad Nacional Autónoma de México, Centro de CienciasC BY-NC-ND license (http://creativecommons.org/licenses/by-nc-nd/4.0/).

ation rate (CIR); Service level differential zone (SLDZ)

aroni, 2009), energy monitoring (Lu & Gungor, 2009), andncreasingly in the fields of optical character recognition OCRKorber, Wattar, & Scholl, 2007) and thermography (Fluke,007).

Thermography measures the temperature of the entire picturehat the charged-coupled device (CCD) has captured. For exam-le the Ti-300 infrared thermography imager manufactured byluke can be applied in a typical industrial solution to monitornd inspect furnaces and boilers wirelessly using its proprietarySD-wireless’, on-board 802.11 g/n interface (Fluke, 2008). Themager provides a real-time video image and thermographicnalysis of the temperature distribution inside the combustionhamber or boiler to Android/Linux nodes. This information isrucial for process control and preventative maintenance pro-edures, but it requires broadband capability for networked

ransmission.

ISM band, OFDM WLANs are the proposed solutionor WSNs that require broadband operation with committed

Aplicadas y Desarrollo Tecnológico. This is an open access article under the

2 lied R

ii(TrE

mpel8iIntG

da2oiq8ccoad

apcuabe

tnSutpto

aIaoau

ir

almmsf

fpbowWf

ta(Rfao

1

m‘Ttctimco

bna

itdTpca

astc

P. van Rhyn, G.P. Hancke / Journal of App

nformation rate and predetermined class of service. Twondustry measurement standards for data throughput, RFC2544Giguère, 2008) and ITU-T Y.1564 (ITU, 2016), are applicable.he terms committed information rate (CIR), excess informationate (EIR) and class of service (CoS) are defined by the Metrothernet Forum (MEF, 2006) and by the ITU-T in Y.1564.

CIR and EIR are specifically defined by Y.1564, in terms ofeasurement philosophy and methodologies, as the acceptable

arameters in the method to determine service performance lev-ls that conform to modern industrial and commercial serviceevel agreements. It therefore appears logical that modern IEEE02.11a/g/n systems that transport TCP/IP Ethernet services forndustrial solutions should be evaluated according to the sameTU-T Y.1564 standards. It should be noted here that 802.11b isot an OFDM-based system, but that 802.11a/g/n OFDM sys-ems are backward compatible with 802.11b (Paulraj, Nabar, &ore, 2003).Multiple-input multiple-output orthogonal frequency-

ivision multiplexing (MIMO-OFDM) wireless technologyppears to fulfil the requirements of industrial WLANs (Giguère,008). MIMO-OFDM is considered a most interesting devel-pment in the field of broadband wireless technology becauset is extremely tolerant to multi-path propagation and fre-uency selective fading. A major advantage of an OFDM-based02.11a/g/n gateway router, for WSN use, is its 3G/LTE Internetonnection capability that facilitates a global communicationonnection (Liu & Li, 2005), seamlessly connection dissimilarperating systems across dissimilar networks. A far-reachingpplication in mind is that Android/Linux-based IWSNs can beeployed, operated and controlled from anywhere on the globe.

Performance gains may be made in OFDM systems by thepplication of more than one antenna at either or both connectingoints of the wireless link (Bolcskei, 2006). Multiple antennasan be used to effect interference cancellation and may also besed in coherent combining techniques to obtain diversity andrray gain. An additional fundamental channel gain is realizedy the use of multiple antennas in a MIMO configuration atither/both sides of the link that improves spectral efficiency.

Performance gains may also be made in OFDM systems byhe application of a higher gain antenna at either or both con-ecting points of the wireless link (Ide, Kingsley, O’Keefe, &aario, 2005). Higher gain antennas like log periodic types aresually highly directional with a beam width of 15◦–20◦ in sec-or specific direction (azimuth). Attention is drawn to verticallyolarized omnidirectional antennas with higher gain, such ashe multiple-element collinear array, (Karlsen, 2009) which willnly operate satisfactorily in the horizontal plane.

Exceptional performance gains may be made by selecting theppropriate technology that is available to the industry (Nationalnstruments, 2016). For example, a trend observed in both theutomotive and military industries is to make use of commercialff-the-shelf (COTS) technology, repackaged to IP65 standards,s opposed to development of turnkey systems from the ground

p.

This approach holds many advantages for industrial electron-cs and informatics practitioners because of time and cost savingselated to the measurement platform. For the same reasons, the

m

wi

esearch and Technology 15 (2017) 1–13

pplication of Android/Linux-based platforms has become pro-ific (National Instruments, 2016). These devices incorporate

ostly 802.11a/g/n compatible WLAN capabilities, as well asultiple connectivity options. Their capabilities to connect dis-

imilar operating systems across dissimilar networks place themunctionally on par with OSI model, level 7 gateways.

The latest commercial gateway/routers operate in both ISMrequency bands of 2.4 GHz and 5.8 GHz in the WLAN, soroviding frequency diversity. Each frequency band is utilizedy two omnidirectional transceivers, hence providing a degreef space diversity, as well as noise cancellation. The gate-ay employs a so-called ‘hotstick’ USB modem to connect theLAN to the Internet service provider, typically in a licenced

requency band such as 3.8 GHz.However, a related problem is to define the degree to which

he network can be improved in terms of its service level as result of performance optimization attempted by the userAkerberg, Gidlund, & Bjorkman, 2011; Bello, Mirabella, &aucea, 2007; Sun, Akyildiz, & Hancke, 2011). Advantages

or industrial electronics and informatics practitioners from thispproach are time and cost savings related to the set-up andperation of the measurement platform.

.1. Research problem, objective and contributions

The results of bit-rate testing by pattern transmission may beisleading if one considers a phenomenon we shall refer to as the

service level differential zone’ (SLDZ) of an OFDM WLAN.he SLDZ of an OFDM WLAN was observed in an indus-

rial environment during preliminary transmission tests that wereonducted by the authors in 2007, in preparation of an engineest bench for the purpose of (destructive) diesel engine test-ng. This form of engine testing requires remote temperatureonitoring of the engine components because of the mechani-

al risks of engine failure which include flying shrapnel and hotil ejection.

It was observed that the OFDM network will maintain broad-andwidth transmission (video and audio soundtrack) with aode, up to a certain distance away from the access point (AP),s the separating range (r) is increased.

At a range rne (never exceeded) from the AP, the signals suddenly compromised with little or no warning by videoransmission quality. The node now has to move back a certainistance towards the AP to rmax in order to regain transmission.his distance moving back was observed to be a considerableercentage of the total apparent transmission range rne. It is thisonsiderable range between rmax and rne that we shall refer tos the SLDZ.

The quality of service (QoS) level of a network cannot beccurately determined when nodes are deployed within (or out-ide) the SLDZ of the network. The SLDZ appears to conformo the magnetic hysteresis phenomenon and underlying physi-al laws that describe it, for example the Jiles and Atherton (JA)

odel of hysteresis (Carpenter, 1991).This paper formulates the hypothesis that connected band-

idth (BW) in Mbps is directly proportional to the power densityn W/m2 of the radio frequency (RF) wave, or conversely to the

lied R

eiobwsRi

rioziic&

mbTrurcpt

tatist

tprbb

semTup

rthcidvt

1

oBtl

tprttc

bfcr

sltr

1

airpst

P

otApsc+

d(

P

i

P. van Rhyn, G.P. Hancke / Journal of App

lectric field strength (E) of the RF wave in V/m, at the receiv-ng node. Connected bandwidth refers to digital bandwidth, aspposed to analogue bandwidth, which is the frequency rangeetween the half power points of an electromagnetic energyave. This implies that the connected BW will decrease at the

ame rate as the power density or the electric field strength of theF wave, until the decline in signal-to-noise ratio (SNR) results

n the termination of streaming wireless communication.Testing the above-mentioned hypothesis positively is a

esearch goal that motivates the application of the well-knownnverse-square law of radiated power, in conjunction with effectsf the phenomenon termed here as the service level differentialone. The consideration of the service level differential zones a novel approach that provides the methods and mathemat-cal building blocks, which are required to solve range, andapacity-related issues of wireless systems (Girs, Uhlemann,

Björkman, 2012).This paper proposes a method to determine the excess infor-

ation rate (EIR) and the committed information rate (CIR)y detection, using the SLDZ dynamic hysteresis phenomenon.his technique is combined with scaled video streaming to rep-

esent the user defined throughput bandwidth. Previous methodssed the file transfer protocol (ftp) method which was not accu-ate because not all the key performance indicators (KPIs) areonsidered namely latency and jitter. The d-CIR (detected CIR)rovides a quality of service reference, which conforms mostlyo the ITU.Y-1564 test philosophy.

This paper contributes the scientific equations and methodshat are required to estimate wireless network capabilities, vari-tions and performance during industrial deployment or fieldests. With these it becomes possible to qualify OFDM WLANsn terms of their service performance according to acceptabletandards such as Y.1564, which is a key requirement for indus-rial applications.

An extended research goal of this study is to devise a methodo compare different OFDM WLAN data links in terms of theirerformance. An acceptable range performance parameter isequired when considering improvement to the link quality,ecause it will determine the maximum extent to which the linkudget can be increased.

The study proposes the typical radio range parameters namelyensitivity and SINAD (signal to noise and distortion ratio) toxpress OFDM WLAN dynamic range performance. These areostly used in the specialized field of radio communications.hese parameters can be derived from the d-CIR and d-EIRsing the SLDZ detection method previously described in thisaper.

A method is exploited to determine the excess informationate (EIR) and the committed information rate (CIR) accordingo the (ITU) and (MEF) by detection, using the SLDZ dynamicysteresis phenomenon, and to equate these to the radio rangeharacteristics of the physical layers under test. This techniques combined with scaled video streaming to represent the user

efined throughput bandwidth. The d-CIR (detected CIR) pro-ides a quality of service reference, which conforms mostly tohe ITU.Y-1564 test philosophy. |

esearch and Technology 15 (2017) 1–13 3

.2. Analytical approach: the free space radiation model

In order to determine the characteristic transmission curvef the wireless link under test, samples of connected bandwidthW are drawn at diminishing range intervals. These samples are

hen processed with MS Excel to produce a logarithmic trendine, depicting connected BW as a function of range in metres.

The stepped curve response of the transmission charac-eristics makes more accurate reading intervals difficult andointless. The connected bandwidth level at each respectiveange section is maintained until the firmware switches overo the next lower (or higher) BW setting. The BW switching cri-eria are not published by the manufacturer so that the effectsan only be observed, but not clearly explained.

A simple solution is a method to smooth the stepped curvesy using logarithmic trend modelling such as Excel or MatLabunctions. More meaningful readings can then be made graphi-ally, and the resulting graph function may be used to calculateange-corresponding BW variations.

The point source radiator or isotropic antenna has to be con-idered in the study of the radiated field from any antenna, whenine-of-sight transmission is considered. Practical antennas andheir performance are often referenced in terms of this basicadiator.

.3. Derivation of the range equations

Energy radiates equally in all directions from the isotropicntenna; therefore, the spherical radiation pattern in any planes circular, for example in the ground plane. The typical isotropicadiator is supplied with a power of P watts at the origin O. Theower flows outward from this point and through the sphericalurface S, which has a radius r. The intensity of radiation, i.e.,he power density PD at point I can be defined as:

Di = P

4πr2 W/m2

This equation illustrates the effects of the inverse square lawf irradiance: at twice the range (2 × r), with P at a constant,he power density PDi decreases by a factor of four or by −6 dB.lso, to obtain the same PDi at double the range (2 × r), theower P would have to increase fourfold or by +6 dB, irre-pective of the number of antennas used (Wolnicki, 2005), asonfirmed by Wavion’s MIMO enterprise AP equipped with six6 dBi antennas.

Maxwell’s theories define the relationship between the powerensity PDi and the E-field and H-field vectors as followsMaxwell, 2011):

Di = Ei × H W/m2

The absolute magnitude of the power density or intensity ofrradiation is thus:

PDi| = EiH = Ei2

120π

4 lied R

i3

dc

L

L

∴

∴

i

t

L

o

r

L

pqak

1

sImata

2

dear

ciwoprep

rcppiattmw8

n2wPcsttmietr

aRsn

uaiti

P. van Rhyn, G.P. Hancke / Journal of App

(where Ei is the electric field strength of irradiance and thempedance of free space is 120� ohm, which is approximately77 �).

According to the inverse square law, the ratio of two powerensities at their respected ranges from the source of radiationan be expressed as the link budget variation, in dB:

ink budget variation (dB) = 10Log

((PDi1/r1)

)((PDi2/r2)

) (1)

Manipulation yields:

ink budget variation (dB) = 10LogPDi1

PDi2

× r2

r1

Since PDi = Ei2/120� (Maxwell, 2011):

Link budget variation (dB) = 10Log

((Ei21/120π

))((

Ei22/120π)) × r2

r1

Link budget variation (dB) = 10Log

(Ei1

Ei2

)2

× r2

r1

But according to the hypothesis, the electric field strength Eis directly proportional to the connected bandwidth BW:

∴ Link budget variation (dB) = 10Log

(BW1

BW2

)2

× r2

r1

∴ Link budget variation (dB) = 20LogBW1

BW2× r2

r1

(2)

When the bandwidth remains a constant (BW1 = BW2), thenhe ratio of the two respective ranges may be expressed as:

ink budget variation (dB) = 20Logr2

r1(3)

(This is the equation proposed for the analytical processingf data samples)

Also, when the range remains a constant (r1 = r2), then theatio of two bandwidths may be expressed as:

ink budget variation (dB) = 20LogBW1

BW2(4)

With Eqs. (3) and (4) it becomes possible to predict through-ut variations in terms of relative RF link budget (in dB), and toualify OFDM WLANs in terms of their service performanceccording to acceptable standards such as Y.1564, which is aey requirement for industrial applications.

.4. Research objective

The main research objective of this study was to devise aimple method to analyze and estimate the performance of anSM-band, OFDM-based, WSN data link. We compare the trans-ission characteristics before and after signal intensification or

ttenuation is configured, and relate the differences in terms ofhe RF link budget (in dB), by the application of Eq. (3). Resultsre then presented for discussion.

e

Ti

esearch and Technology 15 (2017) 1–13

. Methods and test environment

This section describes the experimental methods and the con-itions under which the test results were obtained so that thexperiments may be easily replicated for continuous reviewnd investigation. Then, the analytical approaches are brieflyeviewed before the data analysis method is described.

Testing is in accordance with the neural black-box methodombined with the mathematical approach of functional approx-mation. The digital inputs to the entire system are comparedith the analogue video and audio output obtained in termsf observed quality. The inner hardware of the node or thehysical layer cannot be manipulated by the user, therefore itemains irrelevant to this study. However, different physical lay-rs can now be compared with one another in terms of functionalerformance.

The test topology is shown in Figure 1.1(a)–(c). The testange is shown in Figure 1.2. The typical OFDM test networkonsists of two mobile computers and an ISM 802.11.g accessoint (AP) router. The 802.11g system was used as an exam-le because this system version (g) was already well developedn terms of industrial packaging and antennas/accessories avail-ble. We also used 802.11g as an example because of interest inhe results obtained from single transceiver systems, as opposedo multiple transceivers as in MIMO operation. The fact of the

atter is that the similar results will (predictably) be obtainedhen repeating these tests with any of the IEEE ratified systems02.11a/g/n.

Figure 1.1 depicts the network options for testing. The serverode used option (a) for Experiment 1, option (b) for Experiment

and option (c) for Experiment 3. Typical ISM band OFDM net-ork variations were deployed for streaming video transmission.erformance results were measured and are presented here asonnected bandwidth (BW) in Mbps, as a function of transmis-ion range (r) in metres. The server node is directly connectedo the AP via Cat 5 cable, and also controls the AP adapter set-ings with a browser-based guided user interface (GUI). For theobile node’s adapter, Intel PROSet Wireless adapter software

s used with the mobile node because it is freely available andasy to use. The installation of this additional software enhanceshe functionality and user perception of the measurement expe-ience.

Radiation measurements normally have to be conducted in annechoic chamber or open-space antenna measurement range.esults otherwise obtained will be regarded as conduction mea-

urements rather than radiation measurements, and are thereforeot considered here.

Since the method described here is intended for actual fieldse and refers to the SINAD parameter (ratio of signal to noisend distortion), the results were required to include the effectsntroduced by external noise and distortion. The transmissionests were therefore conducted on an old cricket pitch, as shownn Figure 1.2, providing a relatively large open space with anvenly flat ground plane.

Figure 1.2 shows the plan area of the transmission test range.he approximate latitude and longitude of the test station is

ndicated, as well as the imaginary range line and markers. The

P. van Rhyn, G.P. Hancke / Journal of Applied Research and Technology 15 (2017) 1–13 5

Fig. 1.1. (a), (b) and (c) Experimental test network.

missi

dF

tpaspoin

camc

vTo

2

mt(c1

Fig. 1.2. Trans

-CIR and d-EIR ranges for the three transmission series inigure 3 are also indicated.

Video transmission commences when the mobile node runshe file in the server node’s share folder with Windows’ medialayer. The use of a predetermined video bit-stream is applied in

test detection technique: If the video stream at a specific pre-caled rate has no perceptual impairment in terms of sound oricture, it conforms completely to the agreed CIR. If any noiser video distortion is perceived, it means the key performancendicators (KPIs) are compromised and the video stream doesot conform to the agreed CIR.

The video bit-stream rate was predetermined to represent theommitted information rate (CIR) or agreed QoS throughput, set

t nominally 600 kbps maximum for the experimental require-ent. This was achieved by selecting a suitable length video

lip and scaling the frame rate down in H.264 with Apple’s

t(f

on test range.

ideo editing software until the desired bitrate was obtained.he resulting audio/video interleaved (AVI) file was then savedn the server node in a shared folder.

.1. Experimental methods

The first set of experiments focused on passive testing. Thiseans the determination of the transmission characteristics for a

hroughput of 600 kbps with any variations imposed in a passivei.e., inactive) manner, as compared with the standard low-omplexity (+1 dBi) antenna characteristics as in Experiment.

These characteristics are compared to transmission charac-eristics when using an antenna with a higher gain characteristicWillig, 2008), for example a vertical (+6 dBi) collinear designor 2.4 GHz in Experiment 2. The test was repeated using two

6 lied R

i+

m6pPaowE

2

teA+Be

sdtTEtcr

aambt

rabto(tt

hdrp

2r

et+e

dT

ai

2r

eHaaw

ta

2+

tm8L

mmps

2+

tmtn

ian

wotoo

2

P. van Rhyn, G.P. Hancke / Journal of App

nline resistive attenuators of 6 dB each in conjunction with the1 dBi antennas in Experiment 3.

The second set of experiments focused on active testing. Thiseans that the transmission characteristics for a throughput of

00 kbps were determined for different settings of transmittedower levels, which were electronically controlled by Intel’sROSet Wireless software. The transmission characteristics of

typical 802.11 g single transceiver network operating at 20%f maximum power were determined in Experiment 4. The testas repeated for a power setting of 80% of maximum power inxperiment 5.

.1.1. Experiment 1: Test network service referenceExperiment 1 determined the service reference for the elec-

ronics that was utilized. The electronics used for this testmployed OFDM MIMO technology, manufactured by Belkin,P model 802.11g+ MIMO with fixed transmit power at17 dBm. With the mobile node equipped with the matchingelkin USB g+ MIMO modem, the two transceiver systemnabled a near range data connection rate of 108 Mbps.

The set-up for Experiment 1 is shown in Figure 1(a). Aftertarting the video transmission, the BW was recorded at suitableistance intervals whilst increasing the transmission range tohe point of total signal deterioration, as shown in Figure 2.his point is marked as the detected excess information rate (d-IR) range. The transmission range was then reduced until video

ransmission restarted. This point was marked as the detectedommitted information rate (d-CIR) range. The SLDZ is theange between d-CIR and d-EIR.

Figure 2 contains the data collected during Experiment 1, in graph generated with the aid of Excel. The indicated d-CIRnd d-EIR are for a test bit-rate of 600 kbps that is continuouslyonitored in terms of four key performance indicators namely

andwidth, latency, jitter and errors. Note the range at whichransmission fails (72 m) corresponds to approximately 6 Mbps.

When reducing the transmission range, the transmissionestarts at 55 m at approximately 18 Mbps. Our method to use

trend-line is applied so that more accurate information maye extracted from the stepped transmission curves, or by usinghe equation y = −26 ln(x) + 125.6 also generated with the usef Excel. These stepped curves are due to the manufacturer’sIntel) internal algorithms that are contained in the firmware ofhe OFDM WLAN physical layers, which can be smoothed withhe use of external software packages like Excel.

The SLDZ manifests itself in the same manner as a typicalysteresis curve (Carpenter, 1991); transmission failing sud-enly at d-EIR and restarting back at d-CIR. This pattern willepeat as the mobile node approaches or recedes from the accessoint AP across the SLDZ.

.1.2. Experiment 2: Test network with improved serviceeference

Experiment 2 determined the improvement in the service ref-

rence for the electronics that was utilized. However, instead ofhe standard +1 dBi antennas, the AP was equipped with two6 dBi antennas as shown in Figure 1(b). The mobile node wasquipped with the matching Belkin USB g+ MIMO modem. The

wei

esearch and Technology 15 (2017) 1–13

ata collected during Experiment 2 is depicted in Figure 3 as (b)ransmission +6 dBi.

Figure 3(b) indicates the test results obtained with +6 dBintennas instead of +1 dBi, which means the link budget wasmproved by 5 dB.

.1.3. Experiment 3: Test network with reduced serviceeference

Experiment 3 indicates the reduction in the service refer-nce for the electronics that was utilized, as a control procedure.owever, instead of the standard +1 dBi antennas, the AP was

lso equipped with two 6 dB attenuators fitted in-line with eachntenna as shown in Figure 1.1(c). This means the link budgetas reduced by 6 dB.The set-up for Experiment 3 is shown in Figure 1.1(c) and

he data collected during the experiment is depicted in Figure 3s (c) Transmission −5 dBi.

.1.4. Experiment 4: Test network with node RF power at11 dBm

Experiment 4 determined the service reference for the elec-ronics that was utilized when operating a node at 20% of

aximum power. The electronics used for this test was an02.11g AP router with single transceiver manufactured by D-ink operating at +18 dBm.

The wireless adapter control software on the mobile nodeade it possible to set RF power in 20% decrements from maxi-um (+18 dBm/63 mW) to zero. For this test the mobile node RF

ower was set to 12.6 mW or +11 dBm. The single transceiverystem enabled a near range data connection rate of 54 Mbps.

.1.5. Experiment 5: Test network with node RF power at17 dBm

Experiment 5 determined the service reference for the elec-ronics that was utilized when operating a node at 80% of

aximum power. For this test the mobile node RF power was seto 50 mW or +17 dBm. The single transceiver system enabled aear range data connection rate of 54 Mbps.

The set-up for Experiment 4 and Experiment 5 is essentiallyndicated in Figure 1.1(a) but with a single antenna on the APnd using the internal RF modem situated on-board the mobileode. The results of the active tests are indicated Figure 4.

In Figure 4, the first series presents the transmission resultsith RF power set to +11 dBm (12.6 mW), and the second seriesf results are for RF power set to +17 dBm (50.4 mW). Figure 4hus indicates the transmission range variation when the RFutput power setting is adjusted by +6 dB, which is four timesriginal setting.

.2. Comparative testing

This section briefly reviews the methods and conditions underhich the comparative test results were obtained, so that the

xperiments may easily be replicated for continuous review andnvestigation.

P. van Rhyn, G.P. Hancke / Journal of Applied Research and Technology 15 (2017) 1–13 7

Fig. 2. Transmission characteristics for test network with +1 dBi antennas.

F ennas

OI

EUn8

msficEit

2

teA+Bes

aam

ig. 3. Transmission characteristics for MIMO test network with (a) +1 dBi ant

The test topology is shown in Figure 5(a) and (b). The typicalFDM test network consists of two mobile computers and an

SM 802.11.g access point (AP) router.The server node used option (a) for Experiment 6 and for

xperiment 7. The mobile node in Experiment 6 used an externalSB modem. Nevertheless, a different manufacturer’s mobileode was used for Experiment 7, equipped with an internal02.11a/g/n physical layer, as in Figure 5(b).

Experiment 6 focused on the determination of the trans-ission characteristics for a throughput of 600 kbps with two

tandard low-complexity (+1 dBi) antennas in a MIMO con-guration as depicted in Figure 5(a). These characteristics areompared to the transmission characteristics obtained during

xperiment 7, when using a single transceiver (Txcvr) AP as

ndicated in Figure 5(b). Note that, the RF power level for bothests, were set to equal levels at +17 dBm each.

mNa

, (b) +6 dBi antennas and (c) 6 dB attenuators in-line with +1 dBi antennas.

.2.1. Experiment 6: MIMO test network service referenceExperiment 6 determined the service reference for the elec-

ronics that was utilized. The electronics used for this testmployed OFDM MIMO technology, manufactured by Belkin,P model 802.11g+ MIMO with fixed transmit power at17 dBm. With the mobile node equipped with the matchingelkin USB g+ MIMO modem, the two transceiver systemnabled a near range data connection rate of 108 Mbps. Theet-up for Experiment 6 is shown in Figure 5(a).

Figure 6 contains the data collected during Experiment 6, in graph generated with the aid of Excel. The indicated d-CIRnd d-EIR are for a test bitrate of 600 kbps that is continuouslyonitored at the receiver node in terms of four key perfor-

ance indicators namely bandwidth, latency, jitter and errors.ote the range at which transmission fails (72 m) corresponds to

pproximately 6 Mbps. When reducing the transmission range,

8 P. van Rhyn, G.P. Hancke / Journal of Applied Research and Technology 15 (2017) 1–13

Fig. 4. Single transceiver transmission characteristics with actively controlled transmit power, indicating typical stepped curves for two series.

test n

tmm

2s

etO8

sta

2t

Fig. 5. (a) and (b) Experimental

he transmission restarts at 55 m at approximately 18 Mbps. Aethod to use a trend-line is applied so that more accurate infor-ation may be extracted from the stepped transmission curves.

.2.2. Experiment 7: Single transceiver test networkervice reference

Experiment 7 determined the service reference for the

lectronics that was utilized when operating a link with a single-ransceiver node. The electronics used for this test employedFDM MIMO technology, manufactured by Belkin, AP model02.11 g+ MIMO with fixed transmit power at +17 dBm.

oaA

etwork for comparative testing.

Nonetheless, the mobile node for this test was equipped with aingle-transceiver system that enabled a near range data connec-ion rate of 54 Mbps. The transmission results for Experiment 7re shown in Figure 7.

.3. Analytical approach and methodology to comparativeesting

The analytical approach is mathematically simplified, inrder to accommodate a much wider spectrum of industrialnd scientific users who may not all be expert mathematicians.ll data was first captured into MS Excel, and then processed

P. van Rhyn, G.P. Hancke / Journal of Applied Research and Technology 15 (2017) 1–13 9

Fig. 6. Transmission characteristi

Fig. 7. Transmission characteristics for single transceiver test network, indicat-i

tsytst

FdAdbsb

dt

gt

maai

rynv

3

diwtsat

(cc

3

oobtained with the +6 dBi antennas, thus a total link budget vari-ation of +5 dB.

ng stepped curve with corresponding logarithmic trend-line.

o produce logarithmic trend-line equivalent curves with corre-ponding mathematical expressions. The resulting trend-line’s-intersect was then adjusted until the curve intersected withhe last minimum recorded (70 m) at the distance boundary, ashown in Figure 6 for the MIMO test network and Figure 7 forhe single transceiver.

With the aid of the logarithmic trend graph indicated inigures 6 and 7, the BW values corresponding to the d-CIR and-EIR can be more accurately extracted from the data collected.s indicated, the transmission characteristics were measured inecreasing steps as determined by the wireless adapter’s on-oard firmware and algorithms. The result is a rather erratictepped curve with more than one range value for each BW valueelow 54 Mbps.

According to the trend graph in Figure 6 (which can also be

escribed by y = −26.6ln(x) + 125.6 for the section of interest),he BW value at a range of 55 m (d-CIR) is 19 Mbps. The trend o

cs for MIMO test network.

raph extends in range up to the last measured BW value beforeransmission terminated.

For the section of graph beyond 70 m, the interpolationethod used here to estimate the BW at 72 m (d-EIR) is by

ssuming a straight line graph section between 70 m (9 Mbps)nd 75 m (0 Mbps). If 72.5 m corresponds to 4.5 Mbps, then thenterpolated bandwidth at 72 m is approximately 5 Mbps.

Considering the single transceiver’s characteristics andesulting logarithmic trend graph as depicted in Figure 7,

= −14.17ln(x) + 71.356, the d-CIR range corresponds to a con-ected bandwidth value of 14.5 Mbps and the d-EIR range aalue of 2.5 Mbps.

. Results and data analysis

In the method followed to test the hypothesis, samples wererawn from the data collected during experimentation presentedn Section 2. After starting the video transmission, the BW dataas recorded at suitable distance intervals whilst increasing the

ransmission range to the point of total signal deterioration, ashown in Figure 3. These data samples were then processed withn Eq. (3) derived in the previous Section 1.3, which is based onhe inverse square law of radiated power.

The equation reveals the relationship between range variationr1/r2) and signal strength (dB) when the bandwidth (BW) isonstant (at 600 kbps). Each of the processed results is thenompared with the expected value.

.1. Passive test data analysis

Considering Figures 2 and 3, Table 1 compares the resultsbtained with the standard +1 dBi antennas to the results

Considering Figures 2 and 4, Table 2 compares the resultsbtained with the standard +1 dBi antennas to the results

10 P. van Rhyn, G.P. Hancke / Journal of Applied Research and Technology 15 (2017) 1–13

Table 1Transmission results +1 dBi antenna vs. +6 dBi antenna.

Measured result Calculated resultdB = (20Logr2/r1)

Expected resultdB = (+6 − 1 dB)

Data rate r1 r2

36 25 45 5.11 536 32 57 5.01 536 40 70 4.86 518 45 75 4.43 518 50 90 4.81 518 55 100 5.19 5CIR 57 102 5.05 59 65 115 4.95 59 70 125 5.03 5EIR 72 127 4.93 5Standard deviation 13% –

Table 2Transmission results +1 dBi antenna vs. −6 dB attenuator.

Measured result Calculated resultdB = (20Logr2/r1)

Expected resultdB = (–6 dB)

Data rate r1 r2

36 40 20 −6 −618 50 25 −6 −6CIR 55 27 −6.1 −69 60 30 −6 −69 70 35 −6 −6EIR 72 37 −5.7 −6Standard deviation 5.5% –

Table 3Transmission results +17 dBm rf-out vs. +11 dBm rf-out.

Measured result Calculated resultdB = (20Logr2/r1)

Expected resultdB = (+17 − 11 dB)

Data rate r1 r2

36 17 37 6.7 636 20 40 6 618 25 50 6 6CIR 27 54 6 69 32 65 6.1 69 35 70 6 6EIR 37 73 5.9 6Standard deviation 16.6% –

ob

3

wr(a

Table 4Transmission results +1 dBi antenna vs. +6 dB antenna.

Measured result Calculated resultdB = (20Logr2/r1)

Expected resultdB = (+6 − 1 dB)

Data rate r1 r2

36 36 62 4.72 +518 56 100 5.03 +5CIR 55 102 5.36 +59 70 125 5.03 +5EIR 72 127 4.99 +5S

3

TTb

3

arTs

e9ecr

4

4

MVt

Toette

ombw

By using a video bit stream of a pre-determined rate to specifythe required committed throughput, the service level differential

btained with 6 dB attenuators set fitted in-line, thus a total linkudget variation of −6 dB.

.2. Active test data analysis

Considering Figure 5, Table 3 compares the results obtainedith the RF transmit power set to 50.4 mW (+17 dBm); to the

esults obtained with the RF transmit power set to 12.6 mW+11 dBm). This represents a 6 dB difference in power level, or

1–4 power ratio. z

tandard deviation 2.6% –

.3. Standard deviation

The standard deviation for each of these three series inables 1–3, are respectively estimated at 13%, 4.2% and 16.6%.his is regarded as excessive in the first and last cases and cane attributed to the stepped transmission curves.

.4. Processing data to improve standard deviation

In a second method of data analysis, collected data samplesre processed to produce a logarithmic trend-line, in order toeduce the standard deviation i.e. accuracy of the measurements.he following Figure 8 indicates a logarithmic trend-line for theeries +6 dBi shown in Figure 3(b).

Using the logarithmic trend-line, specific values can bextracted for the respective data rates 36 Mbps, 18 Mbps,

Mbps, d-CIR and d-EIR. Similarly, specific values werextracted from the series in Figure 2, and the two sets of pro-essed values can again be compared in terms of the expectedesults, as shown in Table 4 below.

. Discussion

.1. The hypothesis

The hypothesis suggests that connected bandwidth (BW) inbps is directly proportional to the electric field strength (E) in/m, or to the power density (PD) in W/m2, of the RF wave at

he receiving node.To test the hypothesis, several experiments were conducted.

he hypothesis was confirmed by the experimental resultsbtained and presented in this study, since they correlate withxpected values. The positive testing of the hypothesis suggestshat connected bandwidth BW in Mbps is directly proportionalo the relative signal strength in dBm, which is the preferredxpression of signal power level.

Positive testing of the hypothesis motivates the applicationf the inverse squared law of radiated power (or irradiation) as aathematical tool to estimate information rate variations in ISM

and OFDM wireless Ethernet networks, as applied in industrialireless sensor networks.

one SLDZ can easily be measured and the CIR and EIR for

P. van Rhyn, G.P. Hancke / Journal of Applied Research and Technology 15 (2017) 1–13 11

s for +

eEt(p

wlicaRkmip

4

(amtut61

tbid

S

rtCCtTc

S

FwF

S

∴

∴

ii+eV

Ts

Fig. 8. Transmission characteristic

ach network option can be detected. Once the d-CIR and d-IR transmission ranges for the network options are known,

hese values can be used in conjunction with a simple set of Eqs.3) and (4) to estimate range or bandwidth variations in networkerformance, in terms of relative link budget in dB.

The observation of a video stream in order to evaluate net-ork key performance indicators (KPIs), namely bandwidth,

atency, jitter and errors, provides a fast, intuitive and easilynterpreted method to confirm the CIR under actual operationalonditions. It appears easier to use BW to express relative vari-tions in signal strength in dB, because the measurement ofSS in dBm requires special test equipment and an in-depthnowledge of test and measurement w.r.t. resolution bandwidtheasurements. BW readings may be obtained from on-board

nstrumentation such as Window Task Manager (Networkingane) or PROSet Wireless freeware from Intel.

.2. The digital SINAD ratio

An interesting phenomenon initially observed is that the CIRin the example’s case 600 kbps max) requires a minimum ofpproximately ten times the connected bandwidth, to a maxi-um of approximately thirty times the connected bandwidth,

o reconnect or reset. These limits are determined by the man-facturer’s firmware, situated on the radio chipset. This meanshe 600 kbps CIR fails at a connected bandwidth value below

Mbps, and reconnects or recovers from reset at approximately8 Mbps connected bandwidth.

With specific reference drawn to the observation above,he signal-to-noise-and-distortion ratio (SINAD) is consideredecause it is an indicator of the quality of the signal thats transmitted from a communications device, and is oftenefined as:

INAD = Psignal + Pnoise + Pdistortion

Pnoise + Pdistortion

aTc

6 dBi with logarithmic trend line.

By functional comparison, the d-EIR is the transmissionange at which the signal level of the CIR has deteriorated tohe point that it is overcome by the noise and distortion level.onversely, the d-CIR is the transmission range at which theIR signal level is at a sufficiently higher ratio elevated above

he noise and distortion level for acceptable video transmission.his situation relatively represents the digital SINAD ratio whichan be calculated from the proposed equation:

INAD (dBcir) = 10LogEffective bandwidth

CIR(5)

With reference to the results of the passive tests inigures 2–4, the SINAD for each characteristic was calculatedith values obtained from the Excel trend-line method as inigure 2. For example, for the curve Tx +1 dBi:

INAD (dBcir) = 10LogBWd−CIR − BWd−EIR

CIR(6)

SINAD (dB) = 10Log19 × 106 − 5 × 106

600 × 103

SINAD = 13.35 dBcir

For the curve Tx +6 dBi, using the same method, the SINADs calculated at 13.18 dB, and for the curve Tx −5 dBi the SINADs 15.44 dB. Acceptable results were noted for Tx +1 dBi and Tx6 dBi, since the SINAD required for microwave frequency isxpected to be slightly higher than the 12 dB recommended forHF.A higher than expected SINAD of 15.44 dB was noted for the

x −5 dBi curve to produce acceptable results. The additionalignal level that is required to produce acceptable results is prob-

bly to overcome additional return loss because of mismatch.he mismatch probably occurs as a result of the additionalonnectors introduced to the system by the attenuator pads.

12 P. van Rhyn, G.P. Hancke / Journal of Applied R

Table 5Comparison of experiment 6 – mimo dual transceiver and Experiment 7 – singletransceiver.

MIMOExperiment 6

Single transceiverExperiment 7

SINAD 13.35 dB 13.01 dBSensitivity 19 Mbps for

approx. 13 dB14.5 Mbps forapprox. 13 dB

4

Ftabc

c

S

∴

∴

i1

lf

mtHpaa

5

wc

mtWmnL

W

sttwkpm

trepctla

C

R

A

B

B

C

C

F

F

F

G

G

G

H

I

SINAD atCIR = 600 kbps

SINAD atCIR = 600 kbps

.3. Comparative test results

The sensitivity specification of the first mobile node inigure 5(a) as a radio receiver, for the transmission characteris-

ics shown in Figure 6, can also now be estimated as 19 Mbps for 13.35 dB SINAD at throughput CIR of 600 kbps. (18 Mbps maye used being the closest standard BW value that the networkan connect, for an approximated value of 13 dB SINAD).

For the second example, the SINAD can be calculated for theurve depicted in Figure 7 by the same method:

INAD (dBcir) = 10LogBWdCIR − BWdEIR

CIR(7)

SINAD = 10Log14.5 × 106 − 2.5 × 106

600 × 103

SINAD = 13.01 dBcir

Again, the sensitivity specification of the second mobile noden Figure 5(b) as a radio receiver can now be estimated as4.5 Mbps for a 13 dB SINAD at a throughput CIR of 600 kbps.

The results obtained by the study, for two dissimilar physicalayers transporting the same CoS traffic, can now be tabulatedor brief discussion (Table 5):

It is noted with interest that the single transceiver has auch improved sensitivity rating. This was not clear from ini-

ial measurements, because the same d-CIR range was measured.owever, with the use of the logarithmic trend graph it becameossible to estimate the corresponding data rate with closerccuracy and clearly reveal the difference in physical layer char-cteristic features.

. Conclusion

The study introduces WSNs to ISM-band OFDM WLANs,ith specific reference to thermography and Android-based

omputer technology.A novel method to estimate a specified committed infor-

ation rate in terms of range and link budget is positivelyested against expected values. These capabilities render OFDM

LANs operating within ISM bands an extremely useful com-unications transport medium for industrial wireless sensor

etworks, and specifically for cutting edge Android nodes withinux-based operating systems.

A hypothesis is proposed that connected bandwidth in OFDMLANs is directly proportional to the relative RF signal

I

K

esearch and Technology 15 (2017) 1–13

trength. The concept is mathematically proved by a heuris-ic statistical analysis approach, whereby data collected fromransmission characteristics are compared with expected results,hen the signals are intensified or attenuated to specificallynown values. A method is revealed whereby collected data arerocessed in order to substantially improve the accuracy of theeasurements.Comparative studies by physical experimentation revealed

hat mobile nodes can be individually tested in terms of theirespective SINAD parameters, as well as their sensitivity param-ters, without connecting external test equipment. With thesearameters available, mobile Android/Linux computer nodesan in future be qualified for industrial use in terms of theirhroughput performance, which is a key requirement for serviceevel agreements (SLAs) which incorporate the ITU-T Y.1564nd EtherSAM test methodologies.

onflict of interest

The authors declare no conflict of interest.

eferences

kerberg, J., Gidlund, M., & Bjorkman, M. (2011). Future challenges in wire-less sensor and actuator networks targeting industrial automation. In The9th IEEE International Conference on Industrial Informatics proceedings,INDIN 2011 (pp. 410–415).

ello, L. L., Mirabella, O., & Raucea, A. (2007). Design and implementationof an educational testbed for experiencing with industrial communicationnetworks. IEEE Transactions on Industrial Electronics, 54(6), 3122–3133.

olcskei, H. (2006). MIMO-OFDM wireless systems: Basics, perspectives, andchallenges. IEEE Wireless Communications, 13(4), 31–37.

arpenter, K. H. (1991). A differential equation approach to minor loops in theJiles-Atherton hysteresis model. IEEE Transactions on Magnetics, 27(6),4404–4406.

arullo, A., Corbellini, S., Parvis, M., & Vallan, A. (2009). A wireless sensornetwork for cold-chain monitoring. IEEE Transactions on Instrumentationand Measurement, 58(5), 1405–1411.

lammini, A., Marioli, D., Sisinni, E., & Taroni, A. (2009). Design and imple-mentation of a wireless fieldbus for plastic machineries. IEEE Transactionson Industrial Electronics, 56(3), 747–755.

luke Corporation. (2007). Why thermography is good for your business, Appli-cation Note 2788354, A-EN-N Rev B. December, Publication ID:11226-eng.

luke Corporation. (2008). Inspecting furnaces and boilers’, Application Note2524871, A-EN-N Rev B. January, Publication ID: 11062-eng.

iguère, B. (2008). RFC 2544: How it helps qualify a carrier Ethernet network.EXFO application note No. 183.1AN.

irs, S., Uhlemann, E., & Björkman, M. (2012). The effects of relay behaviorand position in wireless industrial networks. In IEEE International Workshopon Factory Communication Systems Proceedings (pp. 183–190).

ungor, V. C., & Hancke, G. P. (2009). Industrial wireless sensor networks:Challenges, design principles, and technical approaches. IEEE Transactionson Industrial Electronics, 56(10), 4258–4265.

ou, L., & Bergmann, N. W. (2010). System requirements for industrial wirelesssensor networks. In 2010 IEEE Conference on Emerging Technologies andFactory Automation (ETFA) (pp. 1–8).

de, J. M., Kingsley, S. P., O’Keefe, S. G., & Saario, S. A. (2005). A novelwide band antenna for WLAN applications. In 2005 IEEE Antennas andPropagation Society International Symposium (Vol. 4), IEEE (pp. 243–246).

TU-T Y.1564. (2016). Ethernet service activation test methodology.http://www.itu.int/rec/T-REC-Y.1564/en

arlsen, F. (2009). Guidelines to low cost wireless system design. WirelessCommunications,. Nordic VLSI ASA.

lied R

K

L

L

M

M

N

P

S

W

P. van Rhyn, G.P. Hancke / Journal of App

orber, H. J., Wattar, H., & Scholl, G. (2007). Modular wireless real-time sen-sor/actuator network for factory automation applications. IEEE Transactionson Industrial Informatics, 3(2), 111–119.

iu, H., & Li, G. (2005). OFDM-based broadband wireless network – Designand optimization (1st Ed.). ISBN 0-471-72346-0.

u, B., & Gungor, V. C. (2009). Online and remote motor energy monitoringand fault diagnostics using wireless sensor networks. IEEE Transactions onIndustrial Electronics, 56(11), 4651–4659.

axwell, J. C. (2011). A dynamical theory of the electromagnetic field.

pp. 579–584. The Scientific Papers of James Clerk Maxwell. ISBN9781108012256.

etro Ethernet Forum (M.E.F.). (2006). Ethernet services attributes – Phase 2technical specification, November, MEF 10.1.

W

esearch and Technology 15 (2017) 1–13 13

ational Instruments. (2016). What is a wireless sensor network, NI White Paper7142/EN.. www.ni.com/white-paper/7142/en/ (Accessed 10.05.16)

aulraj, A. J., Nabar, R. U., & Gore, D. A. (2003). Introduction to space-timewireless communication. Cambridge University Press.

un, Z., Akyildiz, I. F., & Hancke, G. P. (2011). Capacity and outage analysisof MIMO and cooperative communication systems in underground tunnels.IEEE Transactions on Wireless Communications, 10(11), 3793–3803.

illig, A. (2008). Recent and emerging topics in wireless industrial commu-nications: A selection. IEEE Transactions on Industrial Informatics, 4(2),

102–124.

olnicki, J. (2005). The IEEE 802.16 WiMAX Broadband Wireless Access;Physical layer (PHY), Medium Access Control layer (MAC). Radio ResourceManagement (RRM) Technical Report.